Machine learning in the oil and gas industry

New information technologies – artificial intelligence and, in particular, machine learning as a process of creating self-learning algorithms – can change not only the Internet and the information sphere. Machine learning is being successfully implemented in heavy industry, including the oil and gas industry. Machine learning is not widely used in this industry yet, but potentially such algorithms can be used:

• when selecting ways to increase oil recovery from fields;
• when optimizing oil and gas transportation routes by any means;
• when building logistics for the supply of equipment for oil production;

• to build forecasts for the development of new fields;
• to analyze energy consumption and prices.

The oil and gas industry has several important features that necessitate the search for innovative solutions – the continuity and high complexity of the technological chain, which begins with geological exploration and ends with the delivery of oil and gas to consumers. This is a geographical separation of the place of production and the place of processing, and these places can be separated by tens of thousands of kilometers. The equipment operates in difficult conditions, and even modern geology cannot always predict any factors that will affect the recovery of the field.

For example, a machine learning predictive and incident prevention system for offshore oil and gas wells and pumping units has enabled British Petroleum to reduce its total operating costs by $ 2 million. This is due to the fact that downtime before repairs has been reduced due to the improved distribution of efforts of the repair teams.

General Electric Oil & Gas Corp. has implemented machine learning algorithms to optimize oil and gas equipment diagnostic scheduling, reducing downtime while waiting for service. The sensors that the company installs on technology continuously feed machine learning algorithms with data for automated analysis, and now diagnostics are carried out exactly where problems can be expected.

Proactive troubleshooting reduces accidents, which also increases production and decreases production costs.

An important element of machine learning in the oil and gas industry is the concept of a “digital field”, which is essentially a computer model of a real field, on which a variety of forecasts and models can be worked out. Data for a digital field comes from the production site, for which a well or several wells are supplied with sensors linked into a single system. This approach allows you to increase the level of oil and gas production, reduce costs, predict emergency situations and eliminate their causes. Thanks to this, many oil and gas companies, in particular Shell, Chevron, British Petroleum, reduce the cost of mining by 8-10% and extend the life of equipment by 20%.

Machine learning algorithms can also help identify anomalies in the resource extraction process. Modern oil and gas fields are equipped with a huge number of sensors, but the information that comes from them can only be effectively processed by a self-learning system, it is impossible to do it manually. Artificial intelligence makes it possible to predict possible problems, including those caused by geological anomalies that were not detected in time. The possibility of obtaining a recommendation for setting the production mode is also important. The main task here is to optimize the operating mode of the pumps, which a person simply cannot do, since he does not have or cannot process all the data necessary for this.

And, of course, machine learning helps in matters related to logistics and economics. Thus, the appropriate algorithms allow creating accurate and adequate business models for the production and transportation of energy resources. The oil and gas industry, which is inflexible due to its complex production structure, is becoming more adaptable to fairly volatile oil and gas prices through the use of machine learning. Such approaches reduce losses from overproduction and insufficient storage capacity.

The most global and at the same time the least developed direction is automation in geological exploration – namely, the creation of a system for constructing geological models based on the available big data in this area. This will reduce the uncertainty inherent in geological exploration in general. After all, only about one in two of the exploration wells is eventually developed – it is expected that machine learning will be able to reduce the number of unproductive exploration wells.

AVADA MEDIA develops various software solutions in the field of artificial intelligence and machine learning. We implement projects of all sizes around the world and are ready to offer the creation of machine learning algorithms, including for the oil and gas industry, from exploration and production to the processing and sale of petroleum products and gas.